YouTube was the most frequently cited website in Google AI Overviews across a panel of vitamin and supplement searches, according to new research. The video platform appeared in 186 of 350 AI-generated answers, or 53.1%, making it the only website cited in more than half of the answers collected.

Searcherries followed 50 searches daily from August 30 to September 5, 2026. The questions came from Google Autocomplete and covered topics including hair growth, energy, sleep, anxiety and weight loss. Every search produced an AI Overview, the AI-generated summary that Google displays alongside its search results.

GoodRx was the second most frequently cited website, appearing in 48.3% of answers. Healthline followed at 28.6%, Cleveland Clinic at 28.0%, and the National Institutes of Health domain, nih.gov, at 25.1%. Websites were ranked by the number of answers citing them, with each website counted only once per answer.

YouTube’s presence extended across 44 of the 50 queries. It was particularly prominent in the group beginning with “vitamins for,” appearing in 78.6% of those answers. Across the full panel, YouTube appeared almost twice as often as Healthline.

The research also found that AI Overviews frequently cited pages outside the recorded organic results. Only 32.4% of cited pages appeared in the organic listings collected for the same search; 67.6% did not. Matching by website instead of exact page raised the overlap to 52.5%, showing that some AI citations pointed to a different page on a website already present in organic search.

Individual answers were less consistent than the overall leaderboard. For “supplements for weight loss,” nih.gov appeared every day and YouTube appeared on six of seven days, while other websites moved in and out. By contrast, “supplements to help with glp-1 side effects” returned the same cited pages throughout the week.

About the research

Searcherries monitored a fixed set of 50 Google Autocomplete queries once daily using US English desktop results configured for New York. The analysis covered 350 AI Overviews and 5,036 citation occurrences across 572 distinct pages and 218 websites. Organic overlap was calculated within each search, counting repeated citations to the same page once; the recorded organic lists contained seven to ten results.

The post YouTube Appears in 53% of Google AI Overviews for Vitamin and Supplement Searches appeared first on AI News.

Just over half of American adults, 50.5%, say a partner’s romantic or sexual relationship with an AI can count as cheating.

The figure comes from an AI romance survey of 2,150 U.S. adults, run by AI Girlfriend Coach through SurveyMonkey Audience on 26 and 27 August 2026 and analysed on the 1,709 responses that remained after quality screening removed 441. Respondents were asked how they would regard a partner’s romantic or sexual relationship with an AI, what they think draws people to these products, and whether they could develop genuine feelings for one themselves.

The question is not as strange as it sounds. AI companions flirt, remember what they were told last week and raise it again unprompted, sustain a relationship across months, and simulate intimacy convincingly enough that users describe grief when a model is retired. Unlike pornography or a fictional character, they respond to the specific person using them, and the relationship develops through interaction rather than sitting fixed and identical for everyone who opens it.

AI romance does not fit any existing definition of cheating

Infidelity norms assume a third person who can be confronted, resented, or left over, and AI romance removes that person while keeping most of what surrounds them. There can still be secrecy, emotional attachment, sexual content, and a substantial share of someone’s attention going somewhere their partner cannot see.

Relationship researchers have long treated emotional infidelity as damaging in its own right rather than as a lesser version of the sexual kind, and AI companions land in that first category while generating chat logs and images that resemble the second. Neither of the categories couples usually argue from quite covers it.

Regulators have run into the same gap from another direction. China’s rules on anthropomorphic AI services took effect on 15 July 2026, while California already has companion-chatbot legislation on the books and New York lawmakers passed additional AI-companion legislation in 2026, with much of the regulatory emphasis falling on minors, disclosure and crisis handling. What these products do inside adult partnerships has been left to households to work out.

The survey also undercuts the assumption that this mainly concerns single people. Engaged and married respondents were nearly twice as likely as single ones to have used an AI girlfriend or boyfriend app, 30.7% against 16.3%.

The discomfort runs wider than the cheating label

A further fifth (21.2%) said they would not reach for the word cheating but would still be uncomfortable with a partner doing it. That takes the share applying some kind of limit to 71.7%, against fewer than three in ten (28.3%) who dismissed the question on the basis that an AI is not a real person.

The disagreement is therefore not really about definitions, since a partner does not need to win an argument over what technically constitutes infidelity for the behaviour to cause damage. Roughly seven in ten are placing some kind of limit around AI intimacy whatever noun ends up attached to it.

Within the half who did use the word, a third (33.4%) said any romantic or sexual interaction with an AI already crosses the line, while 17.1% said it becomes cheating only once emotional attachment develops, placing the boundary at the feeling rather than at the software.

The people open to AI feelings judge AI romance more harshly

Among respondents open to developing genuine romantic feelings for an AI, 73.0% said AI romance can count as cheating, against 41.2% of those who said they never could. The gap of nearly thirty-two points runs opposite to the obvious assumption, which is that sceptics would be strictest and anyone open to AI attachment would wave the question through.

Experience produces the same pattern rather than the reverse, with current AI companion users calling it cheating at 80.6%, former users at 60.2%, and people who had never tried one at 44.0%. Familiarity with these products is associated with a harsher judgement rather than a softer one.

Once someone accepts that feelings directed at an AI can be genuine, the distinction between a real and an artificial relationship stops doing much work in deciding whether a boundary has been crossed. The object of the attachment being synthetic stops functioning as a defence, which leaves the people with firsthand experience of these products applying the firmest rules to them.

AI companionship is not primarily about sex

Asked what would be the main reason for using an AI companion, 56.6% chose having someone to talk to or feeling less alone, against 10.3% who chose affection, romance or sex, and loneliness was the leading answer in every subgroup measured.

That makes the cheating discussion more complicated rather than less. Had AI companionship turned out to be primarily sexual it would map onto existing norms about pornography reasonably well, since most couples have some working arrangement there. An ongoing emotional relationship competing for attention, disclosure and intimacy is harder to file, resembling an undisclosed close friendship more than it resembles adult content.

Respondents largely see these products as a response to loneliness rather than to sexual demand, whatever they are being sold as, which places a heavier responsibility on the category than its marketing usually implies.

Couples will have to define these boundaries themselves

Every earlier version of this question was settled privately in the end. Couples negotiate their own terms around pornography, contact with ex-partners, dating apps left installed, how much of a friendship belongs on a private thread, and how close is too close with a colleague. None of those has a universal rule either, but each arrived with decades of accumulated convention to argue from.

AI companions did not. The survey finds opinion split almost exactly down the middle on the word itself, a clear majority saying the behaviour is at least worth discussing, and no established norm for anyone to fall back on. Couples are being asked to legislate for themselves, on a technology most of them have not tried, during the period when the category is growing fastest.

Explicit AI clauses will probably become as ordinary in relationship conversations as deleting the apps became, though not before a good deal of avoidable argument. For anyone using these products in the meantime, the working assumption the data supports is that roughly seven in ten people would place some kind of boundary around it.

AI Girlfriend Coach tests AI companion platforms firsthand and publishes original research on the industry and the people using it. Full methodology, the questionnaire and the excluded responses are published with the survey.

The post 50.5% of Americans Say AI Romance Can Count as Cheating appeared first on AI News.